Epistemology of open source controllers

Why does my car GPS work every time I turn the key? Why does my boat fishing sonar GPS work every time I use it? Don’t get me wrong. I enjoy working with Pixhawk, PX4. QGC, …etc. But it seems inherently unstable. may not be long before the commercial solution is cheap enough and reliable enough to displace all the open source stuff (DJI??). Thoughts?

MowJack

LOL !!!

That is basically the difference between a consumer product and a DIY Project

If you look at GPS data you will see that you get them as fast as on your car or sonar.
The main difference is that we are dealing with controling the vehicle with the GPS data so we want to unsure that the datas are consistent before using them as cannot allow jump in GPS position.

You easily saw that when you car GPS got the wrong direction on a runabout. On car GPS they will always position the vehicle on a the road so is easier for positionning too !

Now, if you also need to compare GPS solutions as generally on drone world people use cheap one’s. Get a good GPS and you will have awesome performances too.
Generally, the issues you are seeing aren’t related to GPS but to the state estimation from the EKF. We are dealing with really cheap componant so, unfortunately that is to expect. Using good product, but expensive ones, will show you that open source flight controller as really great. The best proof is that on ArduPilot we got a lot of partner doing full autonomous flights daily without pilots supervisor !

1 Like

Is there a consumer rover controller that works out of the box?

Yes
https://ardupilot.org/ardupilot/docs/common-rtf.html

Just an observation that it seems like we should get stable product “out of the box”. There must be billions of GPS/accel/mag products in use by auto, phone, recreation segments. So our hardware is good it’s the flight stack that lacks? Again I enjoy pastime. Not arguing. Just trying to sniff out the best products. Thanks for the reply.

MowJack

Honestly that statement is just ridiculous. I appreciate you are not trying to argue but you are saying stupid things using absurd logic premised on incomplete understanding. My car GPS in my very expensive Land Rover totally messed up my navigation from Ithaca to Chicago btw. There are many an expensive smartphones that have poor GPS performance. So what is your point? I literally see no point in your posting. If you want out of the box performance buy Skydio or DJI products. The only point in buying very expensive RTF open source UAVs is if you have a niche application. The aforementioned consumer UAVs are excellent. This is a DIY forum essentially. With enough effort you can get a plane or copter that will perform very well. I have many planes that work flawlessly.

And when you have actually done the work setting one up you will have greater understanding of why I posted this.

Hello Marc,
Thanks for replying to my “Rants and Raves” post.
I’ll ask this: Is there a quality and or performance difference between say an mRobotics brand controller and a generic offshore clone? Same for telemetry? And GPS?

If said US company or any company says “here is our offerings and they are superior reliability and stability. But we hand to lock parts of it down to achieve that stability. So it is somewhat open source”. Would this community embrace those products? Would you embrace them?

MowJack

That is better phrasing of the question. In short yes. A company like mRo is more likely to source high quality components – especially IMU sensors and barometers. The difference may not be immediately apparent to hobbyists, if there is a difference at all. For the small subset of users that are flying expensive payload and doing professional work then it obviously pays to get your open source compatible flight controllers from a reputable manufacturer. If you are flying a foam plane then a $20 F4 board and a $20 cloned Ublox GPS, all purchased from Bangood, will probably work just fine. The community does embrace products from Holybro and Hex that produce equipment that is compatible with Open Source products but whose actual designs are not open sourced. These days I would say not many companies actually release their design files. Why should they?

Lot’s of them are embraced/supported. There are more than are listed here. You can buy a “Open Hardware” Flight Controller made in the USA if you like.

Mitch,
my Arducopter-controlled drone also works every time! I built it roughly a year ago, and after weeding out the minor problems I had at the beginning, mostly related to poor propeller balance, it has behaved very well over close to one year. No instability of any sort.
This is with a Pixhawk clone, and Arducopter 3.6.8, which admittedly is outdated, but since it does everything I need, I have seen no need to update it.
My drone is a quad, equipped with camera both for recording and FPV, gimbal with elevation control, grabber, of course with GPS. Despite my flying location in dense forest, the GPS starts up quickly and reliably, every time. Return-to-home typically has an accuracy withing about 3 meters, sometimes 5 meters. That’s coming down between tall trees, so I find that accuracy pretty amazing, even if I have to make manual corrections on final approach, if I want the drone to land on my 1*1m pad.
If you have any instabilities with the parts you bought, then you need to sort out what is causing them, and fix it. It could be a loose connection, or a faulty module, or… Well, anthing can fail. But it IS possible to make reliable, dependable drones using these inexpensive modules and free software! And that’s pretty amazing!

That is certainly encouraging! I probably underestimated the complexity when heading down the open source path. Maybe a thread on this site with proven configurations would help steer the newbies like me. With parameter files. I think I made it past that point now but that initial help could expand the community. Also would sort people into skill levels. Just a thought.

MowJack

Indeed it might be helpful for beginners to see several proven configurations, as examples. But since this scene is very dynamic, chances are that in most cases it will be impossible to exactly copy any of those proven setups, because some of the components will already have gone obsolete.

To provide one example of a proven configuration (proven by me, that is…), I wrote a web page about my drone:

http://ludens.cl/aeromod/uglybird/uglybird.html

Ardupilot is really well documented, so one can do almost everything with it just by patiently reading and following the provided instructions. Unfortunately this is not the case with many of the hardware components, so one has to search the web and put together the information required to get a given setup running. For me, doing this learning effort was the main point in my drone project. I had the time, and was willing to spend it. But someone who simply needs a drone, NOW, is really better off buying a factory-made product, that comes configured and ready to go, rather than build his own.

Somewhat of a continuation of your thoughtful reply and a reset to the original post: Driving yesterday with collision avoidance made me want the same gps that detected the vehicle drifting in the lane. There must be a vendor or two that supplies the world wide auto industry. We need those and at the same price the automakers pay for for them!

Mowjack

They are the same devices and there is no way they will be priced anywhere near the same. That’s like comparing Walmart to a tiny Mom & Pop shop.

Think the thread has run it’s course.

I’m not sure the thread has run its course, and besides, it’s a rant, not a technical topic… That said, let’s get technical.

First, lane detection is done by cameras/color sensors, not GPS. ArduPilot can be fed obstacle data from similar systems should you want to invest in them.

Second, automotive GPS is no better than anything we have access to install on our vehicles. If the system in your car seems more accurate, that’s because it either leverages your cell phone to increase accuracy via cell tower/wifi positioning and/or uses AI to assume that it’s on the nearest road and traveling in the general direction of your GPS track. Have you ever watched your car’s symbol travel down an exit you haven’t taken or continue down a highway when you’re on an access road? That’s the navigation software making assumptions regarding its position that in most cases results in a perceived high level of accuracy but sometimes fails.

A uBlox F9P or even M8P module almost certainly has more native capabilities than any given automotive GPS in its raw form. ArduPilot’s navigation system is nothing short of impressive on many flight controllers, going from boot to a fully aligned INS with GPS correction in a matter of seconds or maybe a minute or two, which is easily faster (and, with RTK, more accurate) than commercial or even military aircraft GPS/INS alignments (4-9 minutes, typically, sometimes requiring manual starting coordinate entry and/or manual GPS almanac updates).

Rant and rave all you want, but I don’t want to use an automotive navigation system with all of its assumptions and limitations for a mower in a field or a copter doing aerial photography/survey…or…well, just about anything other than a road-going car.

1 Like

Just for fun, if you want industry scale pricing on a discrete Zed-F9P module, you just need 6 months’ lead time and an investment of $34,697.36…and you can have yourself a sweet GPS module for the low, low unit price of $138.79…at which point you need to have a circuit board capable of accepting it and the reflow soldering skills to affix it.

But you can share the discount with 249 of your closest friends :smiley:

This is a meaningless rant. I have a copter with two proper Drotek F9P GPSs. It works great!
I get a reliable fix in 20seconds.

Just get proper F9Ps (with big patch antennas, or even better with helicoidal antennas) mount then on proper masts and it will work as you want today. No need to complain, just need to buy proper a GNSS receiver available today.

And yes, these are better and more expensive than any automotive GPS.

All great replies! Disclosure: I it is just a rant. And rant replies can contain some great big picture info. Funny how rants are sometimes valuable that way.

Shout out to Yuri. I found your YouTube videos before I found this community. And found Yuri after a part time farmer and work chum was talking about RTK in agriculture. Not knowing any thing about micro controllers, GPS, mags, etc. I set about to learn. With a topic like this it’s best to try get the big picture to better judge your entry point.

And… I’m 61 and way behind on the etiquette of blogging, posting, don’t do Facebook. So no malice intended when I post. Just trying to figure it all out.

Thanks again,
MowJack

2 Likes

Rant? more like speculation and a bit of philosophy. Good reading. my 2c
My flight control is more reliable than my car ! go figure, and i got the cheapo 2.8.4 one from china.
As a state engine - building ardu-stuff has so many possibilities. My car has 4 doors 3 pedals and goes forward or backwards. (yep simplified but making a point) I must say im really happy my car doesnt have a couple of hundred parameters to tune - and that my rover does. I think its a sad day when we can tune our cars infinitely and not our automated flight robots/subs/mowers/balloons/trackers…